Abstract:
Impending rollover events are detected based on vehicle roll rate, roll angle and lateral acceleration, and an assessment of the relative polarities of vehicle roll rate and lateral acceleration. When the vehicle roll rate and lateral acceleration are opposite in polarity, the roll rate vs. roll angle thresholds used to distinguish between rollover events and non-rollover events are more sensitive than for conditions for which the vehicle roll rate and lateral acceleration are of the same polarity (64, 66, 68). Additionally, the roll rate vs. roll angle thresholds are adaptively modified based on roll angle and lateral acceleration to provide increased detection sensitivity under conditions that typically precede a rollover event (80), and reduced detection sensitivity under conditions for which a rollover event is unlikely (86).
Abstract:
An iterative process involving both genetic programming (18) and adaptive boosting (22, 24) is used to develop a classification algorithm (24) using a series of training examples. A genetic programming process (18) is embedded within an adaptive boosting loop (22, 24) to develop a strong classifier (24) based on combination of genetically produced classifiers.
Abstract:
A vehicle rollover sensing apparatus (10) and method are provided for detecting an overturn condition of the vehicle. The rollover sensing apparatus (10) includes an angular rate sensor (12) for sensing attitude rate of change of a vehicle and producing an output signal indicative thereof. The rollover sensing apparatus also has an integrator (64) for integrating the sensed attitude rate of change signal over a variable time window (80) and producing an attitude angle. The rollover sensing apparatus further includes deployment logic (66) for comparing the attitude angle and attitude rate of change to a pair of variable threshold values, with a gray-zone (150) that varies based on time, and an output (56) for deploying a vehicle overturn condition signal based on the comparison. Adaptive bias removal and output minimum logic (62) reduces bias and noise associated with the sensed signal.